蔬菜嫁接机器人嫁接苗特征参数的视觉测量方法  

Vision method for measuring grafted seedling properties of vegetable grafted robot

作　　者：张雷[1] 贺虎[1] 武传宇[1] 

机构地区：[1]浙江理工大学机械与自动控制学院,杭州310018

出　　处：《农业工程学报》2015年第9期32-38,共7页Transactions of the Chinese Society of Agricultural Engineering

基　　金：国家自然科学基金资助项目(51175475);浙江省自然科学基金资助项目(LY14E050027)

摘　　要：嫁接用苗的直径、生长点坐标、苗长等特征信息,是判断能否嫁接匹配的有效参数特征,也是迅速获取机器人嫁接位置参数的重要依据。为了提取相关蔬菜嫁接机器人的嫁接用苗特征信息(嫁接苗生长点坐标、砧木苗子叶茎截面的长短轴直径、穗木苗子叶茎截面长短轴直径、砧木苗长度及穗木苗长度等),该文提出了一种图像处理综合算法。该算法首先确定采集后图像的初步目标范围,然后进行预处理,获得单色图像,利用灰度阶跃变化选定兴趣区域,而后对图像进行中值滤波和图像增强;利用基于高斯拟合、求反和基于大津法阈值分割相结合的信息提取方法,获得生长点横坐标,结合形态学开闭处理方法及逻辑搜索运算,引入有效行连续的概念,剔出强光噪声的干扰,获得了所需的各长短轴直径图像坐标;利用标定结果和相机图像到世界坐标转换的对应关系,获得最终各项指标信息。在自行研发的嫁接机器人样机上,以葫芦类砧木苗为试验对象,经过500次试验,与传统手工实际测量值相比,该算法实测值平均误差小于0.0053 mm,直径最大误差小于0.02 mm,从而验证了该算法的可行性和有效性。该算法能在线获取嫁接苗特征信息,满足嫁接实时要求。The diameter of the seedlings, the growing point coordinates and the seedling length are not only key parameters of grafting robot for judging whether it can successfully graft, but also the important basis for the robot to estimate motion parameters and spatial position of an object. With the aim to achieve feature （the coordinate of growth point of grafted seedling, the long and short axis radius of the seedlings of rootstocks’ cotyledon, the long and short axis radius of the seedlings of scion’ cotyledon, the length of the seedlings of rootstocks and the seedlings of scion） measurement of grafted seedling for vegetable grafting robot, an algorithm of integrated technology was proposed based on machine vision and image processing. The several grafting seedling properties can be obtained on-line. The flowchart of the algorithm was given. There were three mainly steps included in the algorithm. Firstly, the image was preprocessed to find the target area by adjusting the camera focal length, setting the relative position between the camera and the seedling. Then, the seedling color image obtained by camera in experiment was transformed into a monochromatic image. Secondly, the target area was segmented by discontinuous gray for building the range of interesting. A Gauss average method was used to extract mid-line for thex-coordinate of the growth point of grafting seedling. An order statistics filtering with a 5×5 median mask was used to reduce the random noise. In order to increase the dynamic range of the gray levels in the seedling image being processed, a contrast stretching transformation was obtained and its the minimum and maximum gray levels was respectively 0.45 and 0.65. Image complement and threshold segmented based on OTSU were acquired. Finally, the algorithm was made with morphological methods and logical calculation to obtain the image coordinates of the long and the short axis radius of the seedlings of rootstock and scion. A manual measurement method was carried before the method based vi

关 键 词：机器人 图像处理 算法 嫁接 蔬菜 机器视觉 特征参数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]